Program Semantics and Classical Logic

Combining our findings, we conclude that

Rm⁺¹ ⊆ [{V(L^t,α) I ∃L0 : L 'C [hXi ^ PK,ι(Pj)∕X,]‰1L⅛L⁰ 'C Pk}

Since, by Lemma 20, L⁰ 'C Pk implies that

[(Xi ^ Piin=I(Pj)/XjL 'C [(Xi ^ Piin=ι(Pj)∕X]n=ιPk

It follows that

R '   ⊆ [{V(L^t,α) I L 'C [(Xi ^ Pi)i=ι(P,)∕X_i]J=ιPk}

⊆ [{V(L^t, a) I L 'C (Xi ^ Piii=ι(Pk)}

We may conclude that S_m Rk' ⊆ S{V(L^t, a) ∣ L 'C (Xi ^ Pii*=₁(P⅛)} for all
k and now reason as follows.

V(((Xi ^ Piin=ι(Pk))^t,α) = (Lemma 8)
V (Pk^t,α[Rω/Xι,...,Rω /Xn]) = (continuity)
^[V(Pk^t,a[R1^m/X1,...,Rn^m/Xn])=^[Rk^m ⊆
mm

[{V(L^t,a) I L 'C (Xi ^ Pi)i=ι(Pk)}

This concludes the proof. □

10 Conclusion and Further Prospects

In this paper we have shown that at least some non-trivial programming
constructs, including recursion, can be provided with a semantics in the
models of ordinary second-order logic, using the technique of translating
programs and correctness statements into that logic as it is commonly done
in Montague Semantics. The programming language which we have treated
was very small of course, but there seem to be no reasons why we should
not be able in principle to extend it with many useful constructs. This is
one way to continue from here, but there are other possibilities as well. We
speculate on two of them.

• In this paper we have only looked at an imperative language, but the
procedural semantics of logic programming languages is usually done in
the denotational framework as well and makes extensive use of fixpoint
constructions. The idea that we could have a logical treatment of the
procedural semantics of these languages is tantalising.

30

<<   27   28   29   30   31   32   33   >>

More intriguing information